Dept. of Mathematics, Yale Univ., 10 Hillhouse Ave., New Haven, CT 06520
Wavelet packet analysis of sound corresponds to an optimized musical transcription in which the sound is organized in different structures (orchestration), each of which is described by a ``musical score.'' This representation is a superposition of wavelet packets (notes) of different duration, pitch, location in time, and amplitude. Fast algorithms permit real-time transcriptions with a minimal number of notes. In this talk we will give a precise definition of these statements as well as applications to audio compression, denoising, recognition, and modeling.